How does access to this work benefit you? Let us know!
131
0
0
Full text
(2) Maturation of Speech Discrimination and Attentional Requirements in Late Childhood by Judith A. Iannotta. A dissertation submitted to the Graduate Faculty in the Program of Speech and Hearing Sciences in partial fulfillment of the requirements of the degree on Doctor of Philosophy, The Graduate School and University Center at the City University of New York. 2015.
(3) © 2015 Judith A. Iannotta All Rights Reserved.
(4) This manuscript has been read and accepted for the Graduate Faculty in Speech-Language-Hearing Sciences in satisfaction of the dissertation requirement for the degree of Doctor of Philosophy.. Valerie Shafer _________________. _________________________________________. Date. Chair of Examining Committee. Klara Marton __________________. __________________________________________. Date. Executive Officer. Brett Martin Hilary Gomes John Preece Supervisory Committee. THE CITY UNIVERSITY OF NEW YORK.
(5) ABSTRACT Maturation of Speech Discrimination and Attentional Requirements in Late Childhood by Judith A. Iannotta Advisor: Valerie Shafer. The ability to perceive speech sounds and contrasts continues to be refined throughout the course of development. While emerging models suggest that development is characterized by shifts from an attentionally demanding mode of processing speech sounds to one that occurs relatively automatically, the specific developmental time-course of these changes remains unclear. The present work reports the findings of two experiments that aimed to provide insights into the time-course by which neural processes underlying speech discrimination in children and adolescents becomes automatic. The experiments used event-related potentials (ERP) measures, with a particular focus on mismatch negativity (MMN) - a developmentally-sensitive index of automatic speech discrimination. The first experiment focused on children ages 6.0-11.9 years old, comparing the MMN responses elicited by English vowel contrasts under differing attentional conditions in an oddball design, to those observed in adults. Specific attentional conditions included: 1) an auditory-attend condition during which listeners silently counted deviants, 2) an auditory-ignore condition in which listeners ignored auditory stimuli while they being required to solve mathematical equations, and 3) an auditory-ignore condition in which listeners ignored auditory stimuli while passively viewing a silent video. Speech perception was hypothesized to be more automatic in children than adults; accordingly, the MMN was predicted to be more sensitive to manipulations of attention in children than in the adults. Consistent with our predictions, attention-related . i .
(6) modulation (i.e., auditory-attend vs. -ignore) of the MMN observed in frontal and central leads was greater in children than adults. Of note, despite obvious differences in the attentional demands of the two ignore conditions (passive, math), the modulation of the MMN produced by the two conditions differed minimally. The second experiment focused on children ages 10.9-16.9 years old, again comparing their responses to adults. We hypothesized that the maturation of speech discrimination processes would still not be complete in this group and thus expected to find of continued attentional dependencies, in comparison to adults Given the relative lack of differences in modulation produced by the two ignore conditions under experiment 1, we refined our paradigm to include two ignore conditions that differ on a specific cognitive construct – working memory demands. This was accomplished by having participants perform either a 0- or 2-back during the auditoryignore conditions. Consistent with our prediction, we found greater attention-related modulation of the MMN in the left inferior and anterior pole regions for the child group relative to adults; analyses treating age as a continuous variable were supportive of such distinctions. We also found evidence of continued age-related difference in the late discriminatory negativity (LDN) and P3b. Together, these two experiments highlight the value of expanding the scope of examination for speech discrimination to consider a broader range of ages than leading models, which tend to emphasize early life. Our findings of continued age-related changes in the MMN, P3b and LDN during adolescence highlight the need to consider the effects of both bottom-up and top-down attentional influences in developmental models of speech perception.. ii .
(7) . Dedication This work is dedicated in loving memory to my father, Patrick J. Iannotta who encouraged me to follow my dreams and follow my bliss. Who shared with me his love of knowledge and showed to me, through his actions, the power of perseverance.. iii .
(8) . Acknowledgements To my mentors and guides, thank you for teaching me and helping me on this journey. To my friends and colleagues, especially Dr. Rebekah Buccheri-Kallas and Emily Zane, Thank you for your guidance, friendship, and refusal to let me run away ! I want to offer a special show of gratitude to my mother Joan, for her unwavering belief in me. To my siblings, for cheering me on and being my best friends. You make my every day beautiful and I am so fortunate to have you in my life. To my Grace Face and Patrick Henry, for bringing pure joy into my life and heart. Every day. And to my MM Thank you for making all of my dreams come true. ILYWAMHAAF. iv .
(9) TABLE OF CONTENTS Title_________________________________________________________Page Abstract_________________________________________________________i. Dedication ______________________________________________________iii. Acknowledgments________________________________________________iv. Table of Contents ________________________________________________v. List of Appendices ________________________________________________vi. List of Tables ___________________________________________________vi. List of Figures___________________________________________________vi. List of Supplementary Figures______________________________________viii. Chapter 1. Introduction_____________________________________________1 1.1 Statement of the Problem________________________________________1 1.2 Learning & Attention in Classic Models of Speech Perception Development_1 1.3 A Primer on Automaticity ________________________________________5 1.4 Towards An Automaticity-based Perspective of Speech Perception _______8 1.5 Probing the Automaticity of Speech Perception with EEG ______________11 1.6 Mapping the Development of Automatic Speech Perception ____________13 1.7 Remaining Obstacles __________________________________________17 1.8 Summary____________________________________________________19 1.9 The Present Study ____________________________________________19 Chapter 2. Experiment 1___________________________________________20 2.1 Methods ____________________________________________________21 2.1.1. Participants________________________________________________22 2.1.2. Stimulus Materials __________________________________________22 2.1.3. Paradigms and Procedures ___________________________________23 2.1.4. Electrophysiological Recordings________________________________23 2.1.5. Experimental Design_________________________________________24 2.1.6. Data Analysis ______________________________________________24 2.2 Results _____________________________________________________25 2.3 Discussion___________________________________________________34 2.4 Limitations of the First Experiment and Considerations________________37. v .
(10) Chapter 3. Experiment 2 __________________________________________38 3.1 Method _____________________________________________________41 3.1.1. Participants________________________________________________41 3.1.2. Stimulus Materials___________________________________________42 3.1.3. Experimental Procedures _____________________________________43 3.1.4. Electrophysiological Recordings________________________________44 3.1.5. Statistical Analysis __________________________________________45 3.2 Results _____________________________________________________51 4. Discussion____________________________________________________61 4.1 General Discussion____________________________________________64 4.2 Maturation of Speech Discrimination ______________________________66 4.3 Assessing Automaticity_________________________________________67 4.4 Nature of the Speech Stimuli ____________________________________68 5. Conclusion ___________________________________________________69 6. Supplementary Figures__________________________________________71 7. Appendix 1 Primary Analysis with Age as a Dimensional Variable_________81 8. Table 1_____________________________________________________ 105 References_____________________________________________________106. vi .
(11) . List of Appendices Appendix 1. Primary Analysis with Age Treated as a Dimensional Variable ______________81. List of Tables Table 1. Waveforms to the standard stimuli in the attend condition at Cz__________________51 Table 2. Repeated measure ANOVAs for the 100-200ms, 200-300ms, and 300-400ms time windows____________________________________________________________________105. List of Figures Figure 1a. Children Grand Means to standards and deviants averaged across all conditions at C3, Cz, C4, F3, Fz, and F4. ____________________________________________________29 Figure 1b. Adult Grand Means to standards and deviants averaged across all conditions at C3, Cz, C4, F3, Fz, and F4. _________________________________________________________30 Figure 2a. Children’s subtraction waves (deviant minus standard) at C3, Cz & C4 for attend, passive-ignore and math-ignore conditions. _________________________________________31 Figure 2b. Children’s subtraction waves (deviant minus standard) at F3, Fz & F4 for the attend, passive-ignore and math-ignore conditions.__________________________________________32 Figure 2c. Adult subtraction waves (deviant minus standard) at C3, Cz & C4 for the attend, passive-ignore and math-ignore conditions.__________________________________________33 Figure 2d. Adult subtraction waves (deviant minus standard) at F3, Fz & F4 for the attend, passive-ignore and math-ignore conditions.__________________________________________34 Figure 3. First component from the PCA for 100-400msec time window. __________________47 Figure 4. Second component from the PCA for 100-400msec time window. _______________48 Figure 5. First component from the PCA for 450-600msec time window. __________________49 Figure 6. Second component from the PCA for 450-600 msec time window. _______________50 Figure 7a. Waveforms of Stadnard Stimui in the Attend Condition._______________________53 Figure 7b. Plot of GFP, Mastoids and Fz with selected time-windows (100 and 450-600ms ___54. vii .
(12) Figure 8. Correlation between GFP and age under each of the attentional conditions (attend, 0-back, 2-back). _______________________________________________________________55 Figure 9. Child vs. adult responses for attend condition.________________________________56 Figure 10. Child vs. adult responses for 0-back condition_______________________________57 Figure 11. Child vs. adult responses for 2-back condition. ______________________________58 Figure 12. Condition by age-group interaction for mean amplitude in the 100-400msec time window at LI_________________________________________________________________ 59 Figure 13. Child vs. adult responses for attend condition at RP__________________________ 61 Figure 14. Mean amplitude for the left inferior lead for the attend and 2-back conditions in the 100-400 time window.__________________________________________________________ 62 viii. List of Supplementary Figures Figure 15. Children’s MMN waveforms at C3, C4 and Cz in the Attend, 0-Back and 2-Back conditions____________________________________________________________________72 Figure 16. Children’s MMN waveforms at F3, F4 and Fz in the Attend, 0-Back and 2-Back conditions____________________________________________________________________73 Figure 17. Adult’s MMN waveforms at C3, C4 and Cz in the Attend, 0-Back and 2-Back conditions____________________________________________________________________74 Figure 18. Adult’s MMN waveforms at F3, F4 and Fz in the Attend, 0-Back and 2-Back conditions____________________________________________________________________75 Figure 19. Children’s MMN at Fz, Left Mastoid and Right Mastoid in the Attend condition___ 76 Figure 20. Adult’s MMN at Fz, Left Mastoid and Right Mastoid in the Attend condition______76 Figure 21. Children’s MMN at Fz, Left Mastoid and Right Mastoid in the Attend condition___ 77 Figure 22. Adult’s MMN at Fz, Left Mastoid and Right Mastoid in the Attend condition_____ 77 Figure 23. Children’s MMN at Fz, Left Mastoid and Right Mastoid in the 2-Back condition___78 Figure 24. Adult’s MMN at Fz, Left Mastoid and Right Mastoid in the 2-Back condition_____78 Figure 25. Grand Means Standards and Deviants in the Attend condition at C3, C4, Cz, F3, F4 and Fz ______________________________________________________________________79 viii .
(13) Figure 26. Grand Means Standards and Deviants in the 0-Back condition at C3, C4, Cz, F3, F4 and Fz_______________________________________________________________________80 Figure 27. Grand Means Standards and Deviants in the 2-Back condition at C3, C4, Cz, F3, F4 and Fz_______________________________________________________________________81 Figure 28. Standard vs. Deviant Responses for Attend condition with standard error at C3, C4 and Cz__________________________________________________________________________83 Figure 29. Standard vs. Deviant Responses for Attend condition with standard error at F3, F4 and Fz__________________________________________________________________________84 Figure 30. Standard vs. Deviant Responses for 0-Back condition with standard error at C3, C4 and Cz ______________________________________________________________________85 Figure 31. Standard vs. Deviant Responses for 0-Back condition with standard error at F3, F4 and Fz__________________________________________________________________________86 Figure 32. Standard vs. Deviant Responses for 2-Back condition with standard error at C3, C4 and Cz_______________________________________________________________________87 Figure 33. Standard vs. Deviant Responses for 2-Back condition with standard error at F3, F4 and Fz__________________________________________________________________________88 Figure 34. Difference waves in the attend condition by age at F3, C3, Fz, Cz, F4 and C4_____89 Figure 35. Grand Means difference waves in 3 conditions at F3, C3, Fz, Cz, F4 and C4_______90 Figure 36. Difference waves in three attentional conditions at Fz and Cz___________________91 Figure 37. Difference waves in three attentional conditions at F3 and C3__________________92 Figure 38. Difference waves in three attentional conditions at F4 and C4__________________93 Figure 39. Grand Means difference waves at F3, C3, Fz, Cz, F4 and C4 with Mastoids_______94 Figure 40. Condition by Time interaction at C3 in the 100-200 msec time window.__________95 Figure 41. Condition by Time interaction at C4 and Cz in the 300-400 msec time window_____97 Figure 42. Children and Adult MMN waveforms in Attend condition at C3,C4 and Cz_______98 Figure 43. Children and Adult MMN waveforms in Attend condition at F3, F4 and Fz_______98 Figure 44. Children and Adult MMN waveforms in 0-Back condition at C3, C4 and Cz_______99. ix .
(14) Figure 45. Children and Adult MMN waveforms in 0-Back condition at F3, F4 and Fz_______99 Figure 46. Children and Adult MMN waveforms in 2-Back condition at C3, C4 and Cz______101 Figure 47. Children and Adult MMN waveforms in 2-Back condition at F3, F4 and Fz______101 Figure 48. Condition by Time interaction shown at F3 and C3 in the 100-200 msec window__102 Figure 49. Time by Group interaction shown at Fz in the 200-300 msec time window_______103 Figure 50. Condition by Time interaction at C3 and C4 in the 200-300 msec time window___103 Figure 51. Condition by Time by Age-group at F3 in the 200-300 msec time window_______104 Figure 52. Condition by Time by Age-group at Fz, Cz, C3, C4. In the 300-400ms window___105. x .
(15) Chapter 1. Introduction 1.1 Statement of the Problem Despite decades of work, the field of speech science remains without a consensus model that explains speech perception and its development throughout childhood and adolescence . While the specific conceptualizations and details of the many models proposed over the years can differ notably, there are two key questions that most attempt to address. First, how learning shapes innate (e.g., genetically determined) auditory perception of speech, and secondly, whether speech perception at the various developmental stages depends on attention as opposed to being subconscious or automatic processing. In the following sections, we provide an overview of key insights from classic models of speech perception development; then we describe an emerging model, which draws upon well-established theories of automaticity and is amenable to behavioral and electrophysiological examination via manipulations of stimulus complexity and task demands. We review experimental support for this model, as well as gaps in the extant literature and next steps towards a more robust characterization of speech perception development. 1.2 Learning & Attention in Classic Models of Speech Perception Development One prominent model of speech perception is the Word Recognition and Phonetic Structure Acquisition model (WRAPSA) by Jusczyk (1997). This model details a sequence of skills and abilities acquired by infants to explain the development of speech perception in the first year of life. Central to this is the notion that infants innately possess “auditory analyzers” that are capable of encoding most acoustic details from the speech signal. This initial “default” analysis is thought to be broadly tuned and lacking language-specific refinement in early infancy which could explain how infants are capable of detecting sounds of many languages (Werker & Tees,. 1 .
(16) 1984) but fail to recognize subtle but meaningful acoustic differences that adults are capable of distinguishing (Aslin, Pisoni, Hennessy & Perey, 1981). Balancing the concept of innate abilities with learned ones, WRAPSA proposes a structure of the subconscious mechanisms needed for perceptual learning. This includes an experiencedependent “cue weighting” system and a pattern recognition system that, together, account for the attunement of infants to sounds of their ambient language (Jusczyk, 1995; Kuhl, 2004; Werker & Tees, 1984). The term “weight” reflects the relative amount of perceptual attention directed toward specific acoustic information (Nittrouer,2002; pp. 718). The cue weighting system is thought to statistically analyze incoming signals, weight the information and develop weighting “schemes” that will be used to emphasize and enhance the detection of features critical for word distinction. These features are then given greater weight in the perceptual system, leading to a refinement of the analyzers and thus shaping future perception. Werker & Tees’ 1984 Hindi study demonstrated experientially-related refinement by showing that very young infants initially showed sensitivity to both dental /d/ and retroflex /D/ regardless of their ambient language (English or Hindi). By the end of the first year of life however, the sensitivity faded for those whose language (English) considered the contrast to be allophonic but was retained for those whose language (Hindi) recognized it as a meaningful distinction. Complementing the weighting system is the pattern recognition system, which is employed to extracts higher order patterns from inputs and compares them with word templates and exemplars that are stored in the lexicon. This process is strongly related to language experience, as templates are strengthened based on repeated presentations of language input (Jusczyk, 1999; Mattys & Jusczyk, 2001). Once a template is established, subsequent input that. 2 .
(17) matches an exemplar on critical levels is recognized as the same form or “object”, while those that do not will be perceived as a different form or tagged as lacking meaning. A limitation of the WRAPSA model is that it does not explicitly address the attentional requirements for each of the systems included. Given the innate nature of input analyzers, they can be considered to act independent of attention as it is suggested that a child does not intentionally attend to these features. Similarly, the cue weighting system also can be conceptualized as acting without attention, as this system facilitates the identification of speech sounds, which can serve to activate or capture attention. Regarding pattern recognition, while Jusczyk does not specifically state that attention is required, he does suggest that pattern detection is greatly enhanced with attentional focus; he also posited that listeners learn to develop attentional routines to facilitate detection of probabilistic information about sound patterns and likely word forms. Another highly influential model is the Processing Rich Information from Multidimensional Interactive Representations model (PRIMIR) by Werker and Curtin (2005). Like WRAPSA, PRIMIR identifies several components or “planes” of perceptual development, including probabilistic learning, similarity detection, and experiential influences. However, unlike WRAPSA, it does not suggest that any of these planes are present at birth. Instead, the model posits that the components needed for effective perception are the result of a statistical learning process, in which biases are established across the different planes based on regularities in language input. Learning initially focuses on general perception of auditory information, and then is refined to establish phonemic representations, with specificity for sounds from the ambient language emerging over time. The general statistical learning mechanisms specified by PRIMR provide layered simultaneous analysis of acoustic information, prosodic features, segmentation of. 3 .
(18) the speech signal, syllable extraction, phoneme establishment, and storage of word forms, with each having influence over the evolution of the others by enhancing or damping signal saliency. The PRIMR model emphasizes the role of attention in facilitating learning processes. Attentional processes are guided by three dynamic filters that are posited to work and evolve together to develop and refine language-specific speech perception abilities for later language learning. The filters include initial biases of the child, the developmental level, and requirements of the specific language task at hand. The filters interact with one another to direct attention to different cues based upon maturation and task-demands. Similar to WRAPSA and PRIMIR, the Native-Language Magnet Theory (NLM) (Kuhl, 1993; Kuhl & Iverson, 1995), sought to address speech perception development by merging innate abilities and language experience in the formation of perception, through a more distinct framework. The NLM theory postulates that all objects in the perceptual space are analyzed and deemed either relevant or irrelevant to the ambient language experience; attention is not considered to be a necessity for this process. Unlike WRAPSA, which posits that listeners store specific instances of what they hear, NLM suggests that listeners store relevant contrasting objects as phonetic prototypes, each with a specific “neighborhood of influence”. At the center of each phonemic category is the object prototype, which is deemed to be the “best exemplar” and acts as a magnet for other exemplars that closely conform to its relevant features. As experience with language increases, there is a learned refinement of the magnet effect, allowing for more accurate category attraction and specific category development. Non-prototypical objects are not acted upon by the magnet effect and are therefore not perceived as an exemplar of that specific target, creating the potential for the formation of a new phenotypic category as repeated instances occur.. 4 .
(19) Unlike the NLM model, Best’s Perceptual Assimilation Model (PAM) (1995) conceptualizes speech perception to be a highly active process, requiring the listener to seek out information about vocal tract gestures that form specific phonetic contrasts. This view suggests that perception is a purposeful activity by which the listener must attend to a speech signal to gain information about vocal tract positions in order for effective perception and eventual production of the sound(s). One’s perceptual goal of understanding these gestures focuses attention to auditory objects of interest that will be selected form the speech signal and subjected to further analysis. This model is similar to both Fowler’s Direct Realist Theory of Speech Perception (1986) and Lieberman & Mattingly’s (1986) Motor Speech Theory of Speech Perception insofar as they all suggest that speech events are perceived as being articulatory in nature rather than acoustic or auditory. An implicit assumption of these models is that speech perception is attentionally dependent, as a listener must attend to the utterance’s form in order to approximate the physical gestures. From this brief review, the complexity of speech perception, and its development is evident. While each of the models proposed a unique set of organizing principles and specific mechanisms for learning, they tended to vary with respect to their specification of attention; some appear to suggest that attention is beneficial though not necessary (e.g., WRAPSA, NLM), while others have more obvious dependency (e.g., PAM). None of these models provide for a comprehensive or quantifiable understanding of the interplay of attention and learning throughout the course of development. 1.3 A Primer on Automaticity Rooted in cognitive psychology, information processing theory attempts to explain how humans are able to learn and master sophisticated skills by inferring how changes in theoretical mental. 5 .
(20) structures result in the ability to learn new information and execute new behaviors (Eggen & Kauchack, 2007). Central to information processing models is the notion that in order to master a sophisticated skill, an individual must first master the incremental sub-processes that support it. More specifically, the theory attempts to explain how input from the environment goes through the processes of perception, attention, and storage (Eggen & Kauchack, 2007). Focused on fundamental principles, information processing theory has been applied to the study of learning and behavior in a range of areas, including psychology (Forgas & George, 2001) linguistics (Nyikos & Oxford, 1993; McLaughlin, Rossman & McLeod, 1983), technology (Hann, Hui, Lee & Png, 2007), business management (Roger, Miller & Judge, 1999), education (Mayer, 1996), and literacy (LaBerge & Samuels, 1974). LaBerge & Samuels (1974) proposed a highly influential information-processing model of complex skill learning focused on reading. They outlined a series of processing steps involving visual, phonological and episodic memory (e.g., combining letters, mapping graphemes to phoneme, transforming phonemes to semantic representations), which combine together to support the transformation of visual stimuli into semantically meaningful information. Assuming that each of the processing steps is the product of learning, they argued that reading fluency depends not only on the accuracy of the steps carried out, but their “automaticity”. A central tenet of their model was that the automation of lower-order operations or skills is necessary to allow for the allocation of time and resources to more sophisticated, higher-order comprehension skills. For more than three decades, the concept of automaticity has remained a central focus in cognitive psychology, as researchers have worked to operationalize it and apply it to the broader domain of skills learning, beyond reading. Posner and Snyder (1975) highlighted that in order for. 6 .
(21) a process to be automatic, it should occur with or without intention or awareness. Automatic processes are thought to result from being highly over-practiced and do not, on their own, result in the learning and storage of novel information (Shiffrin and Schneider, 1977). Once automatized, there will be little change with increasing age or changes in mental state; Bargh (1989) went so far as referring to them as “chronically accessible”Viewing automaticity in terms of a memory encoding process, Hasher and Zacks (1978) suggested that the encoding of aspects or attributes of inputs for the process must be automatic; they also reinforced the notion that process must be efficiently performed incidentally or intentionally with little to no improvement made with intention or practice. Importantly, Posner and Snyder, 1975 argued that a truly automatic process should not be vulnerable to interference from simultaneous processes once the process has begun. Such processes require minimal cognitive resources, do not require conscious monitoring (Bargh, 1992) and are not vulnerable to interference from other ongoing operations, unless the concurrent process is competing for input or output channels (Bargh,1994). As such, an automatic process can actually become a source of interference when not relevant to task performance (e.g., the Stroop Task). Bargh (1994) attempted to summarize the literature by putting forth a set of criteria upon which the automaticity of behaviors could be judged, which includes: awareness, intentionality, efficiency and controllability. Not surprisingly, the concept of automaticity has emerged in the realm of speech perception - the first step in spoken communication. Understanding how speech perception becomes automatic begins with explaining the underlying perceptual skills necessary for negotiating complex speech signals in a continuous signal (e.g. establishment of phonemes and their boundaries) so that automatic retrieval of word patterns in speech streams can evolve. 1.4 Towards An Automaticity-based Perspective of Speech Perception. 7 .
(22) While a number of studies claim that speech perception is automatic (Johnson & Ralston, 1994; see Shriffen & Schneider, 1984), they do not provide insights into how automaticity develops. As highlighted by the WRAPSA model, which focuses on the first years of life, the elements for speech perception may be present at birth, but automatic processing is not. Instead, children are thought to learn to automatically attend to relevant contrasts through experience with the attentional process itself being affected by language learning. Detection of language-specific contrasts therefore, becomes automatic only after some level of learning has taken place. This is in line with Posner who characterized learning as a highly attention-dependent process in which information must be attended to and prioritized before subsequent memory formation, or learning, can occur. In recent years, the Automatic Selective Perception (ASP) model of speech perception (Strange, 2010; Strange & Shafer, 2008) has emerged as a solution to the challenges of operationalizing and quantifying interplay of attention and learning in speech perception, and its development. The ASP attempts to explain the refinement of language perception in terms of the reorganization of attention to language-specific contrasts. Within this framework, speech perception is thought to be a purposeful activity, in which individuals attend to an incoming speech signal in order to perceive intended messages. After extensive experience attending to sounds of an ambient language, individuals weight acoustic information and form Selective Perception Routines (SPRs). Once established, listeners can detect language-specific sequences, automatically. According to the ASP model, inexperienced language learners, including children, must purposefully allocate attention to speech signals until SPRs are formed and perception is automatized. It is at this point that attention is no longer required for the detection of language-. 8 .
(23) specific objects by the SPR. Frequent repeated exposure will strengthen and refine the SPRs leading to over-learning of an object (Crick and Koch, 1990) and eventually to a reciprocal relationship of increased SPR strength and decreased need for attention. SPR “strength”, for the purpose of this study, will refer to the perceptual salience of an auditory signal. Unlike auditory salience, which refers to the size of the physical differences between phonetic contrasts, perceptual salience refers to the physiological and behavioral responses to language-specific contrasts that occur as a function of linguistic experience (see Strange, 2010 for a review). Highly perceptually salient auditory objects attract attention and eventually, do not require focused attention for their perception (Koch, 2004). Speech perception studies show that both age and language experience contribute to the perceptual salience of phonetic contrasts (Nittrouer & Miller, 1997; Nittrouer, 2002; Shafer, Yu, Datta, 2010; Sundara, Polka, Genesee, 2006). Additionally, Sundara, Polka, and Genesee (2006) found that four year-olds discriminate the English contrast /d- ð/ better than 10-12 month-old children, but poorer than adults and that participants who lacked meaningful experience with the contrast showed poorer discrimination overall regardless of age. Nittrouer and colleagues also provide evidence of developmental changes in perception as they found that children aged 4-8 years-old do not make use of acoustic features the same way adults do when identifying phonemes in a speech signal (Nittrouer and Miller 1997;Nittrouer, 2002). When examining weighting routines during a fricative judgment task, 4-7 year-old children and adults were asked to identify fricatives as being [s] or [ʃ] when followed by an [a] or [u]. An age-related trend was found in which children assigned more weight to the vocalic formant transitions and less to fricative noise when compared to adults. A gradual and reciprocal change in this weighting strategy was found as the age of the children increased, revealing developmental perceptual. 9 .
(24) changes throughout childhood that have not reached adult-like status by age 7 or 8 (Nittrouer and Miller, 1997; Nittrouer, 2002). It has been suggested that perceptual attention is initially focused on formant transitions that provide information about vocal tract changes. As age and experience increased, the focus then shifts to more subtle acoustic properties that do not involve spectral change, such as silent gaps that specify periods of vocal-tract closure; or periods of stable spectral information that specify place of consonantal constrictions (Nittrouer, 2002). The notion that language experience, rather than developmental stage alone, shapes perceptual routines is supported by cross-language findings. For example, adult native Japanese speakers were found to have difficulty categorizing and discriminating the English [r-l] distinction, since it is not phonemically contrastive in their native language (Strange, 1995; Strange, 2010). Such difficulties are not present in Japanese infants, who show good perceptual discrimination of this contrast at 6 to 8 months, though intriguingly not at 10 to12 months of age. Overall, these findings indicate a decline in discriminative sensitivity to this non-native contrast (Tsushima, Takizawa, Sasaki, Shiraki, Nishi, Morio, Menyuk & Best, 1994). Mounting evidence shows that experiential influences (Best, 1995; Hisagi, Shafer, Strange & Sussman, 2010; Jusczyk, 1999; Mattys & Jusczyk, 2001; Strange, 1995; Starnge & Shafer, 2008; Werk & Tees, 1984) attentional processes(Gomes, Molholm, Ritter, Kurtzberg, Cowan & Vaughan, 2000; Hisagi & Strange, 2011; Krause, 1992; Kraus 1993; Sussman, Ritter & Vaughan, 1998; Hisagi, Shafer, Strange & Sussman, 2010 ) and maturational factors (Shafer, Yu & Datta, 2011) affect speech perception, leading to increased interest in developing a more complete understanding of speech perception development considering these factors, as well as their interactions. A key challenge to such pursuits is the need for objective measures of speech processing that can be measured without confound throughout development. While behavioral. 10 .
(25) observations can prove useful in this regard (e.g., comparison of performance outcomes under different attentional conditions), they are not without limitation. In particular, it can be difficult to compare, or equate, behavioral measurements across ages (Bishop, Hardiman, & Barry, 2011) and clinical populations. Additionally, behavioral indices may lack the depth required to capture the complexity of neural phenomena underlying speech processing. Fortunately, electrophysiological tools have emerged as a means of providing objective characterizations of component processes underlying speech processing, and appear to be able to successfully index SPR development. 1.5 Probing the Automaticity of Speech Perception with Event-Relate Potentials Electroencephalography (EEG) is used to non-invasively examine electrical activity of large groups of neurons in the brain. EEG responses are electrical deflections (in the range of microvolts) that occur in response to a stimulus and are detected by surface-electrodes at the scalp. The selective averaging of EEG responses across events of a given type results in eventrelated potentials (ERPs), which reveal time-locked neural processing (Duncan, Barry, Connolly, Fischer, Michie, Näätänen, Polich, Reinvang, Van Petten, 2009; Klix, Näätänen, & Zimmer 1985; Luck, 2005; Näätänen, 1990). Of particular relevance to speech perception development is the Mismatch Negativity (MMN) (Näätänen, 1990), an ERP component that is thought to reflect automatic detection of changes in the auditory signal. The MMN is elicited when a frequently occurring auditory standard is interrupted by an infrequently occurring deviant (Luck, 2005; Näätänen, 1990; Näätänen, Tervaniemi, Sussman, Paavilainen, Winkler, 2001), and is typically described in terms of latency and amplitude. Latency refers to the period of time between the presentation of the deviant stimulus and the onset of the MMN, while amplitude refers to the peak difference between the standard and deviant stimulus (Amenedo & Escera, 2000). The MMN is commonly. 11 .
(26) examined using a classic oddball paradigm, in which a repeated sequence of standard stimuli is infrequently interspersed with deviant stimuli. Participants are typically instructed to ignore auditory stimuli and focus on some other activity, such as watching a silent move, reading a book, or performing a visual task (Muller-Grass, Macdonald, Schröger, Sculthorpe, Campbell, 2007; Muller-Gass, Stelmack, Campbell, 2005; Naatanen, Paavilainen, Rinne, & Alho, 2007; Winkler, Karmos, Näätänen, 1996). Central to its utility in the study of automaticity, the MMN is believed to be the product of a pre-attentive, automatic detection system (Naatanen, 1992; Naatanen, 1990; Naatanen, Gaillard, & Mantysalo, 1978; Sculthorpe, Collin, Campbell, 2008). In considering the significance of the MMN, two competing interpretations have emerged. First, the model adjustment hypothesis postulates that the MMN is an error detection signal that results from a deviation in an established auditory standard. Once regularity in a signal is detected, a predictive model of future inputs emerges. Any unexpected violation of the anticipated standard generates a mismatch response, which triggers online adjustments of the model. In contrast, the adaptation hypothesis argues that the MMN is actually reflects local neural adaptation to the standard stimulus, which causes an attenuation and delay of the obligatory N1 response associated with early auditory processing. It is this N1 differential that the adaptation hypothesis suggests is reflected in the MMN. While the exact origins of the MMN are not definitively known, the EEG, MEG and fMRI literatures appear to be converging on two principle generators – one located in supratemporal cortices and the other in right prefrontal cortex. The respective role of the two generators can be explained in terms of the model adjustment mode. The temporal lobe based generator is thought to be responsible for the automatic detection of deviations between an incoming auditory stimulus and either memory traces of preceding stimuli, or predictions of future stimuli based upon past. 12 .
(27) The higher order prefrontal generator is believed to be responsible for the redirection of attention to an auditory stimulus when the lower-order temporal generator detects a deviation. Importantly, while the MMN is thought to be independent of attention, the ability to discriminate sound contrasts is not necessarily so. It has been reported that that MMN elicitation is only significant when the standard and deviant stimuli are behaviorally discriminable (Martin, Kurtzberg & Stapells, 1999; Kraus et al. 1996). However, studies have shown that with repeated exposure, difficult to discriminate, unfamiliar auditory stimuli can eventually evoke an MMN (Näätänen, Schroger, Karakas, Tervaniemi, Paavilainen, 1993) suggesting learning or sharpening of the sensory system. 1.6 Mapping the Development of Automatic Speech Perception The potential value of the MMN for efforts to study the interface between attention and the development of automaticity comes from findings that attentional processes can enhance the MMN (Datta, Shafer, Morr, Kurtzberg, & Schwartz, 2010). For example, a study by Gomes et al. (2000) compared children (ages: 8-12 years old) and adults with respect to attentional modulation of the MMN, using a paradigm in which a standard tone (1000Hz) and three deviants (easy: 1500Hz; medium: 1200 Hz and difficult: 1050Hz) were presented in an oddball design under three conditions of attentional focus (ignore, attend, ignore again). While adults exhibited robust MMN’s for all conditions, with minimal evidence of attentional modulation, the MMN’s in children were significantly modulated by attention for difficult-to-detect deviants. The presence of significant attentional effects in children, but not adults, supports the claim that perception for some auditory contrasts is automatic in adulthood but not yet in childhood. Datta et al. (2010) also provided evidence of attentional modulation of the MMN in children, with peak responses. 13 .
(28) occurring 50 ms earlier in the attend condition for easy-to-detect vowel contrasts, than in the ignore condition. In an effort to provide a more comprehensive characterization of speech perception development, Shafer and colleagues examined the mismatch responses (MMRs) in children between 2 months and 11 years old, finding the MMN peak shifted earlier by approximately 11milliseconds (ms) per year (Morr, Shafer, Kreuzer, Kurtzberg, 2002; Shafer, Morr, Kreuzer, Kurtzberg, 2000). At age 11, adult-like MMN latencies had not been reached, suggesting the protracted nature of speech processing refinement. In addition, a clear MMN was not observed to the more difficult tone contrast before four years of age (although a positive mismatch response was observed). Similarly, MMN was not clearly present until four years of age to a vowel contrast (Shafer, et al., 2010; 2011). Not surprisingly, these various findings have led several studies to consider the potential value of the MMN in identifying and tracking neurodevelopmental disturbances related to language. While the primary focus of our effort was to map the development of speech perception using the MMN, a number of other ERP components have been shown to be developmentally sensitive and relevant to speech perception. Here, we provide a brief overview of these components; additionally, secondary analyses in the proposed experiments will reference developmental changes seen in these indices, when relevant. P100 (P1). The P100 is an obligatory ERP component that occurs in response to the detection of an auditory stimulus (Näätänen and Picton, 1987), which can be modulated by attention and is developmentally sensitive (Näätänen et al., 1978; Picton and Hillyard, 1974). The component typically manifests as a positive deflection observed between 80 and 140 ms poststimulus. Evidence for attentional modulation of the P100 comes from studies such as that of 14 .
(29) Mangun & Hillyard (1990), who found that when identical stimuli were presented to either an attended or unattended channel, that the P100 in response to the attended channel was larger. From a developmental perspective, the P100 is not detectable early in life. The component initially manifests around 6 months of life, though as the P150, which persists into early childhood (approximately 6 years old). The P100 observed in children is thought to develop in the adult P1 with a steady decrease in latency and amplitude until 20 years of age (Ponton, Eggermont, Kwong and Don, 2000). There is some lack of consensus regarding the developmental phenomenology of the P100, as some have suggested that age-related differences in the waveform may be less related to P1 maturation, and more reflective of amplitude increases in the subsequent N1 (Ponton et al, 2000). N2b. The N2b, a subcomponent of the N200, is a negative deflection occurring between 180 and 325 milliseconds after presentation of the deviant stimulus (Patel & Azzam, 2005). In contrast to the MMN (a.k.a. the N2a), which is considered to be an index of automatic processing, the N2b is thought to index voluntary processing. This N2b component is elicited when individuals detect a deviation of an infrequent or low-frequency stimulus that is being attended to and believed to reflect detection of mismatch between the actual stimulus and the expected stimulus. The N2b is commonly observed for deviations in auditory, semantic, and orthographic information when the specific dimension is being attended to in a selective attention oddball paradigm. The greater the difference between a standard and deviant stimulus, the larger the N2b response observed. A distinguishing property between the N2b and the MMN, is that the MMN exhibits polarity reversals in the mastoid lead, but the N2b does not (Näätänen et al., 1993). As recently reviewed by Patel and Azzam (2005), age effects for the N2b have been reported throughout childhood into adulthood. For example, in a visual color deviation task, decreases in. 15 .
(30) response time, error rates and in N2b latency were shown with increasing age, from 7-24 years old (Van Der Stelt, Smulders, 1998); this was interpreted as being suggestive of increasing cognitive and visual discrimination with development. At the other end of the lifecycle, increases in N2b latency are observed in the elderly, possibly reflecting the compromises in attentional control that are characteristic of aging. P300: There are two P300 subcomponents, the P3a and the P3b. The labels P300 or P3 are commonly used to refer to the P3b, while the P3a requires specific designation. The P3a is a positive going waveform occurring around 250-300ms after stimulus presentation, which is thought to reflect an involuntary shift to novel changes in the environment (Polich, 2003). In contrast, the P3b (henceforth referred to as P3) is a large positive-going wave that peaks around 300-500 ms after stimulus presentation. While the P3a is elicited by infrequently occurring deviants in unattended conditions, the P3 is thought to be elicited by infrequently occurring stimuli that are relevant to an intentional/focused task. There is a direct relationship between how improbable the attended target is and the magnitude of the P3 response, with more rare targets eliciting bigger P3 responses (Donchin, 1981; Escera et al., 2000; Gumenyuk, Korzyukov, Alho, Escera, Schroger, Ilmoniemi, Naatanen,2001.). Morphologically, the P3 tends to be relatively broad waveform, while the P3a is notably more narrow. P3a amplitudes tend to be maximal over frontal/central sites on the scalp, such as Fz or Cz, while P3 amplitudes are generally greater at sites like Pz (Comerchero, Polich,1999). Of note, the P3a is thought to reflect more than simple deviance detection. Rather, it is thought to index further evaluation of novelty to determine if it is of relevance to task performance. This theory is supported by studies that report increased response times after a stimulus elicited a P3a, suggesting that the deviant triggered an involuntary attention switch, and evaluation of the stimulus to determine the relevance of the deviance for. 16 .
(31) current task performance (Escera, Alho, Schroger, Winkler, 2000; Woods, 1992). The greater the deviance, the larger the P3a amplitude (Yago et al, 2001). Developmentally, the P3 tends to increase throughout childhood into adulthood, with decreases emerging in advanced aging. LDN. The late difference negativity (LDN), also referred to as the ‘late MMN’ (Neuhoff, Bruder, Bartling, Warnke, Remschmidt, Müller-Myhsok, Schult-Körne, 2012), is a negative component occurring between 300-550 milliseconds after the onset of rare stimulus (Bishop, Hardiman & Barry, 2011). The LDN tends to be more prominent for speech sounds than nonspeech sounds, and tends to be larger in children than adults. Inspired by their findings that the LDN is larger in small deviant stimuli than large deviant stimuli, Bishop et al. (2011) noted that the LDN should not be viewed as simply a later version of the MMN. Instead, they argued that it appears to reflect additional processing demands associated with salient features of a stimulus that are harder to detect; this may also explain why children, who are less experience with speech perception, exhibit larger LDN responses than adults. 1.7. Remaining Obstacles While MMN-based approaches to probing automaticity in speech perception have gained significant popularity, there is a continued need for methodological refinement. A number of studies have failed to produce attentional modulation of auditory discrimination as indexed by the MMN, raising concerns about sensitivities to the specific conditions and stimuli employed. For example, Shafer, Morr, Datta, Kurtzberg, & Schwartz (2005) found no evidence of attentional modulation of the MMN in 8-10 year old children in an experiment using [I] and [e] vowel contrast; interestingly, the same laboratory later found evidence of attentional modulation when longer duration versions of these vowels (Datta, et al., 2010) were employed. Disparities in findings across studies suggest the complexity of the process(es) indexed by the MMN. Although. 17 .
(32) defined based upon stimulus deviance, the MMN is believed to reflect the hierarchical relationship between top-down intentions and bottom-up perceptions (Garrido, Kilner, Stephan, & Friston, 2009; Grimm & Schroger, 2007). A number of experimental factors have been identified across studies, which may account for some of the variation in findings. In particular, stimulus complexity tends to vary on a variety of dimensions from one laboratory or study to the next (e.g. tones vs. syllables, synthetic vs. natural speech, easy vs. difficult contrasts). Familiarity with the stimuli and contrasts employed can affect the strength of bottom-up processes and thus their susceptibility to top-down modulation. Additionally, task instructions vary substantially, with some studies requiring participants to detect single-stimulus deviants, while others require more complex pattern/rule detection across multiple stimuli. Most germane to the present work, is the fact that the vast majority of studies make use of a “passive-ignore” condition as the low demand condition in their study of attentional modulation. This can be particularly problematic, as there is no way to ensure that participants are actively allocating their attention away from the stimulus; such a situation can be especially problematic in younger populations, who are generally less prone or able to comply with instructions. Beyond the challenge of how to optimize MMN-based approaches to ensure reliability and reproducibility across studies, an important gap remains in the study of early to middle adolescence, where the auditory processes indexed by the MMN is thought to reach adult levels. This is reflective of the tendency of studies to compare early and middle childhood to adulthood. Few studies have mapped out the MMN to speech contrasts during adolescence when it is expected to reach adult levels. Those that have, tended to rely on synthetic stimuli and traditional “passive” ignore conditions, raising concerns about their generalizability and completeness.. 18 .
(33) 1.8 Summary In summary, although encouraging, with respect to validation of automaticity-based models of speech perception, current efforts to map the trajectory of speech perception development via the MMN are incomplete for a number of reasons. First, they tend to make use of suboptimal paradigms that do not allow for careful control or quantification of attentional demands in the manipulations used to probe automaticity; they may not vary attentional demands sufficiently (thereby decreasing sensitivity to age-related changes). Additionally, while a number of MMNbased studies have examined changes in speech perception from early to late childhood, transitions from adolescence to adulthood are yet to be comprehensively examined. An additional limitation of previous studies is that most have used synthetic speech stimuli rather than natural which limits the generalizability of findings as synthetic stimuli lack ecological validity, and are not processed equivalently to natural speech tokens (Nusbaum, Dedina & Pisoni, 1984; Pisoni, 1981). 1.9 The Present Study The present study includes two experiments that aim to: 1) use MMN to map the development of automatic speech perception for natural stimuli from late childhood into midadolescence, when adult levels are expected to be achieved, and 2) refine current methodologies for probing the automaticity of speech perception in child and adolescent populations, using mismatch negativity response. As later sections will describe in greater detail, experiment 2 builds on the insights obtained from experiment 1 to further refine the manipulation of attentional demands in the ignore condition using the N-back task that is employed to pull attention away from the speech signal.. 19 .
(34) Additionally, experiment 2 hones in in adolescence, to provide a more thorough examination of the bridge between childhood and adult levels of speech perception. Chapter 2: Experiment 1 The primary goal of the first study was to examine the maturational trajectory of the MMN to a subtle vowel contrast under carefully controlled attentional conditions. Central to this goal was the inclusion of an attend condition as well as both an active ignore (i.e., ignore the auditory stimuli while mentally solving math equations) and passive ignore (i.e., ignore the stimuli while passively viewing a silent movie) condition to more carefully control for attention. In particular, the active ignore allowed us to ensure that task instructions were followed. The inclusion of two ignore conditions was intended to elucidate a major potential confound in the literature. Additionally, we included a number of improvements as follows: 1) usage of natural speech stimuli to provide greater ecological validity; 2) use of an active ignore task where performance could be quantified and, thus, allow us to assess direction of attentional focus. Given that prior work suggests that the MMN is not adult-like by 11 years of age, we also included an adolescent age range (10-12 years) to determine whether there is mature vowel processing at these ages. We predicted that adults would show a robust MMN in all conditions and that children would show attentional modulation evidenced by a reduced MMN in the ignore conditions with the most dramatic reduction occurring in the math condition. It was hypothesized that the math condition will pull attention away from the auditory signal to a greater extent than the passiveignore and that this would be reflected in the MMN. 2.1 Methods 2.1.1 Participants. 20 .
(35) A total of 34 individuals participated in the first experiment, 8 of whom were children (ages: 6-12) and 11 were adults (ages: 18-40). Among these individuals, 14 were excluded either due to behavioral performance outside of our criteria on the math condition, due to EEG-related artifacts or technical issues in data collection or processing. Two participant groups were established, one consisting of adults (n=11; ages: 18-40) and the other consisting of children (n=8; ages: 6-12). An additional 6 children were excluded from the study because of high artifact, 6 because of poor behavioral accuracy and 10 because of corrupt data files. The child group was composed of an equal number of children ages 6-8 and 10-12 years. Participants were recruited from the greater New York area via flyers and online ads (Craigslist.org & Backpage.org). Prior to participation, we received written consent from each participant’s parent as well as assent from the participant; this procedure is in accordance with The Graduate Center’s IRB requirements and regulations. Inclusion/Exclusion Criteria. All participants were monolingual American-English speakers that resided in monolingual households. Developmental and language history questionnaires, as well as parental interview, were used to exclude any individuals with a history of speech, language, psychiatric or developmental impairments. Additionally, the following formal testing was employed to ensure that language and auditory function were within normal limits: Clinical th. Evaluation of Language Fundamentals, 4 edition (CELF-4) (Concepts and Following Directions, and Recalling Sentences, Understanding Spoken Paragraphs, and Phonological Awareness), puretone audiometric screening (25 dB HL at 500, 1000, 2000, and 4000 Hz). Individuals were excluded if they have any of the following conditions as per caregiver report: emotional or behavioral disturbances, cognitive delay, motor deficits, or neurological signs including seizure disorders or use of seizure medications. Participant histories were negative for hearing. 21 .
(36) impairment, neurological impairment or psychiatric disorder. A small monetary payment of $25 was provided for participation in the study.. 2.1.2. Stimuli Materials Natural speech sequences /æpəә/ and /ɑpəә/ ([æ] as in vowel sound in “cap” and [ɑ] as in the vowel sound in “cop”) served as the stimuli. Three tokens of each syllable type, /æpəә/ and /ɑpəә/, produced by a single male speaker, were used to provide a more naturalistic listening condition. The token durations for /æpəә/, measured in milliseconds (ms) are 434ms, 453ms, 433ms with a mean of 440ms and a standard deviation (SD) of 11.27. The /ɑpəә/ tokens are 400ms, 380ms, and 420ms, with a mean of 400ms and a SD of 20. The stimuli were presented free field over two loudspeakers, situated approximately 1m from participants at a comfortable listening level of 65 dB SPL with an interstimulus interval (ISI) of 300ms and a stimulus onset asynchrony (SOA) of 800ms. Natural stimuli were employed because they are more ecologically valid than synthetic speech and provided natural variability across tokens. Native listeners have shown excellent discrimination for these stimulus contrasts but poorer perception has been found for non-native listeners (Shafer, Strange, Ito, Gilichinskaya, Rosas & Kresh, 2011). The order of the stimuli was pseudo-randomized for each subject with presentation adhering to the restriction that deviants must be separated by at least three standards, with an overall probability of 20% deviance and 80%. 136 deviants were delivered for each condition. 2.1.3 Paradigms and Procedures In the attend condition, a behavioral discrimination task was conducted during which participants were asked to count the deviant /ɑpəә/ tokens in six 3-minute blocks. After each. 22 .
(37) block, the stimuli were stopped and the examiner documented the number of deviants the participant reported hearing. In the math condition, accuracy was recorded to ensure that the participants were engaged in the task and that the task was of an appropriate difficulty level. Math lists were assigned prior to the study based on performance on a brief pre-test. Performance of 75-90% guided selection of the math list to ensure a task that was challenging enough to require attentional focus but was not too difficult as to result in disengagement. Scores higher than 90% on a pre-test were thought to reflect that the list was too easy and thus would not be considered challenging enough capture attention. In the passive-ignore condition, participants were instructed to attend to a silent movie and ignore the auditory stimuli. 2.1.4. Electrophysiological Recordings A 65-channel Geodesic net, that included two electrooculargraphy (EOG) electrodes, was placed on the participant’s head. Net electrodes made contact with the scalp via saline soaked sponges. The EEG was amplified using the EGI Geodesic Amplifiers, which included an online bandpass filter (0.1 to 40 Hz filter). NetStation version 4.1.2 was used to record the data at a sampling rate of 250 Hz per channel (continuous mode) for later off-line processing. The continuous EEG was processed off-line, using a low-pass filter of 20 Hz, and segmented into epochs with an analysis time of 200 ms prestimulus baseline to 800 ms poststimulus. The data were then baseline corrected and examined for artifact using Netstation software version 4.1.2. Epochs with excessive artifact (i.e., those with differential average amplitude greater than ±70 uV on more than 5 channels) were excluded from the average. Channels identified as bad on more than 20% of the trials were replaced using a spline interpolation algorithm from surrounding channels. ERP averages were then calculated for each stimulus type (standard, deviant) and baseline corrected using the 100-msec pre- stimulus activity. For each participant, we calculated. 23 .
(38) difference waveforms for the examination of the MMN, by subtracting the standard waveform from the deviant waveform. 2.1.5. Experimental Design Participants were seated in a sound-attenuating booth and instructed to listen to a series of natural speech sequences presented in three different attention conditions (passive-ignore, mathignore, and auditory-attend) in a classic oddball paradigm. The two ignore conditions required participants to ignore auditory stimuli while they either watched a silent video (passive-ignore) or conducted a challenging mathematical task (math-ignore) in which they were required to mentally solve mathematical equations and indicate whether the solution provided was correct via a response box (Yes/No response). The auditory-attend condition required participants to actively attend to the stimuli and mentally count the number of deviant sounds presented. Participants reported the number of deviants at the end of a block. Inclusion of the two separate ignoreconditions enabled the examination of the effects of instructing participants to passively ignore stimuli versus the effects of directing their attentional focus away from the signal to a challenging competing task. EEG recordings were obtained throughout performance of the task. 2.1.6. Data Analysis Identification of Time Windows. Time windows selected for analysis were identified after visually inspecting the waveforms. Time-points of interest were identified and analysis was conducted to examine them. Site Selection for Analysis. Six frontocentral sites were selected for analysis. Frontocentral sites were selected because research has consistently demonstrated the largest MMNs at these locations. These sites included Geodesic site 13 (F3) and site 17 (C3), site 62 (F4) and site 54 (C4) and two midline sites, site 4 (FZ) and the vertex (CZ).. 24 .
(39) Statistical Analysis. The primary goal of our analysis was to determine differences between children and adults with respect to the impact of attentional condition on the MMN. In order to accomplish this, at each electrode site, we made use of a 3-way ANOVA (implemented in SPSS) that consisted of the following factors: group (children, adult), attention condition (attend, passive-ignore, passive-math), and time (five 20-second intervals between 100-200ms). Given potential age- and attention-related variations in the time characteristics of the MMN, we repeated our analyses for two additional time intervals (200-300ms, 300-400ms). We made use of a standard p < 0.05 alpha criteria for significance, and employed the Greenhouse Geisser correction to account for potential violations of sphericity. The Helmert contrast implemented in SPSS was used for post-hoc comparisons; this contrast compares the first level of the factor the remaining factors, then the second level to the remaining factors and so on (e.g., for attentional condition: attend vs, [ignore-math, ignore-passive], ignore-math vs. ignore-passive). 2.2 RESULTS Prior to testing for age- and attention-related differences in the MMN, we first inspected the mean responses to standard and deviant trials to ensure the appropriateness of our measurements. Figures 1a and 1b, show the grand mean ERP waveforms for standard and deviant trials for the two age groups, respectively. Figure 2a and 2b display the subtraction waveforms at the six sites selected for statistical analysis for children. Figures 2c and 2d display the subtraction waveforms at the six sites selected for statistical analysis for adults. Across all sites, ANOVA revealed a significant main effect of time in one or more of the time intervals examined (100-180; 200-280; 300-380).. 25 .
(40) Of note, C4 was characterized by a later response than the other sites (F3, FZ, F4, C3, CZ), only showing a significant effect of time in the 300-380 millisecond time-range. Significant interactions observed in each of the three time windows are reported below. Early interval 100-200 ms In the 100-200 millisecond time window, a three-way interaction of group X condition X time was significant at Fz (F(8, 128)=2.72; p < 0.04) and C4 (F(8, 128)= 2.613; p < 0.044). In order to further investigate the group X condition X time at Fz, we first visually inspected our findings and carried out post-hoc analyses. Consistent with the presence of a 3-way interaction, the magnitude of attentional modulation attend vs. ignore [passive, math] appeared to increase with time in children, but not adults. The helmert post-hoc contrast supported this notion. Specifically, the magnitude of group-differences in attentional modulation was significantly smaller at earlier intervals, ms 100 and 200, relative to later time-intervals (100 ms: F(4,64)= 7.32, p < .016; 200 ms: F(4,64)=5.26; p < 0.036). No other post-hocs for this interaction achieved significance. Next, we visually inspected the three-way interaction at C4, finding that children, but not adults, showed evidence of attentional modulation of the MMN. Specifically, adults show comparable MMN amplitude for the three attentional conditions, from 100-200 ms; in contrast, children exhibited no MMN for the attend condition and MMN negativity for the two ignore conditions. Post-hocs only supported the presence of a group difference in the MMN in the comparison of the two ignore conditions at 120 ms versus later time intervals (F(1,16)=6.076; p < 0.025), and 160 ms versus 180 msec (F(1,16)= 4.328; p < 0.054). Beyond the two 3-way interactions noted above, a condition X time interaction was found at C4 (F(8, 128)= 8.392; p< 0.00001) with a trend noted at F4 (F(8,128)= 2.279; p < 0.085). A main. 26 .
(41) effect of time was found at Fz (F(4,64) =6.50; p< 0.006), F4 (F(4,64)=7.443 ; p < 0.001) and Cz (F(4,64)=3.176 ; p< 0.055). Post hoc showed difference between 100 and later (F=5.553; p< .032) and 160-180 (F=3.870; p< .067). Middle interval 200-300 ms A three-way interaction of condition X group X time was significant at F4 (F(8,128)= 4.881; p < 0.002). Visual inspection of the three-way interaction at F4 and post-hoc tests suggested that group differences in the impact of attentional modulation changed with time. Adults appeared to have an MMN response in the 200-300 ms interval only for the attend condition, while children had an MMN response to both the attend and ignore conditions (albeit somewhat smaller for the ignore condition). Helmert-based post hoc tests confirmed the presence of significant group differences in the impact of attentional modulation (attend vs. ignore) on the MMN, with significant effects being found in the comparison at each time interval (200 vs. later [F(1,16)=8.197; p < .011], 220 vs. later, F(1,16)=14.097; p < .002], 240 vs. later [F(1,16)= 7.045; p < .017] and 260 compared to 280 [F(1,16)= 4.118; p < .059]). The two ignore conditions did not differ significantly. A two-way interaction of condition X time was found at C4 (F(8,128) =4.979; p < 0.007), which was the only site that did not reveal a main effect of time (F3(4,64) [F(4,64)= 9.124; p < 0.001], Fz [F(4,64)= 17.422; p < 0.00001], F4 [F(4,64)=11.425; p <0.001], C3 [F(4,64)= 8.923; p < 0.002], and Cz [F(4,64)=5.244 ; p < 0.023]). For the condition X time interaction at C4 (F(8, 127) = 4.979; p < 0.007), a less prominent response was noted for the math condition than the ignore-passive and attend conditions. In support of this observation, post hocs showed differences between the two ignore conditions at 200 versus later (F(1,16)=20.605; p < 0.00001), 220 vs later. 27 .
(42) (F(1,16)=7.703; p < 0.0014), 240 vs. later (F(1,16)=3.527; p < .079) and 260 vs 280 (F(1,16)=3.527; p < .005). Late interval, 300-400 ms A condition X time interaction was found for C3 (F(8, 128)=3.565; p < 0.022), Cz (F(8, 128)=3.897; p < 0.015) and C4 (F(8,128)=4.032; p < 0.006). A significant time X group interaction was found for C3 (F(4, 64)= 4.523; p < 0.022) and a main effect of time was found for Fz (F(4, 64)= 6.94; p < 0.007), F4(F(4, 64)= 4.392; p < 0.019), and C4(F(4, 64)= 4.270; p < 0.012). Regarding the condition X time interaction, inspection of C3 suggested that the MMN in the attend condition was returning to baseline (i.e., becoming less negative) more rapidly than the two ignore conditions. This observation was supported by post-hocs comparing the attend vs. ignore conditions at 300 vs. later (F(1,16)= 10.433; p < .005), 320 vs. later (F(1,16)= 7.452; p < .015) and 360 vs. 380 (F(1,16)=4.198; p < .057). Cz showed differences between attend and ignore conditions at the first (F(1,16)=10.817; p < .005) and second (F(1,16)=5.984; p < .026) time intervals, with the math condition being more negative in the early, but not at later intervals. Finally, the significant time X group interaction appeared to reflect the children having continued negativity in this time window, while the adults had more subtle variation with time. Consistent with this observation, post hoc contrasts for 300 vs. later (F(1,16)=6.531; p < .021) and 360 vs. 380 (F(1,16)=6.483; p < .022) were significant.. 28 .
(43) Figures from Experiment 1.. Figure 1a. Children’s Grand Means to standards and deviants averaged across all conditions at C3, Cz, C4, F3, Fz, and F4.. 29 .
(44) . Figure 1b. Adult Grand Means to standards and deviants averaged across all conditions at C3, Cz, C4, F3, Fz, and F4.. 30 .
(45) . Figure 2a. Children’s subtraction waves (deviant minus standard) at C3, Cz & C4 for the attend, passive-ignore and math-ignore conditions. .. 31 .
(46) . Figure 2b. Children’s subtraction waves (deviant minus standard) at F3, Fz & F4 for the attend, passive-ignore and math-ignore conditions.. 32 .
(47) . Figure 2c. Adult subtraction waves (deviant minus standard) at C3, Cz & C4 for the attend, passive-ignore and math-ignore conditions.. 33 .
(48) . Figure 2d. Adult subtraction waves (deviant minus standard) at F3, Fz & F4 for the attend, passive-ignore and math-ignore conditions. 2.3 DISCUSSION The primary goal of experiment 1 was to examine the development of automaticity in speech perception, as indexed by the MMN, using a cross-sectional sample of children ages 6-14 year old children, with an adult sample (18-40 years) included as a reference. Here, we reported preliminary results from the first phase of data collection, in which we verified our ability to. 34 .
Related documents